A low complexity coding and decoding strategy for the quadratic Gaussian CEO problem

We consider the quadratic Gaussian CEO problem, where the goal is to estimate a measure based on several Gaussian noisy observations which must be encoded and sent to a centralized receiver using limited transmission rate. For real applications, besides minimizing the average distortion, given the transmission rate, it is important to take into account memory and processing constraints. Considering these motivations, we present a low complexity coding and decoding strategy, which exploits the correlation between the measurements to reduce the number of bits to be transmitted by refining the output of the quantization stage. The CEO makes an estimate using a decoder based on a process similar to majority voting. We derive explicit expression for the CEO׳s error probability and compare numerical simulations with known achievability results and bounds.